| 龚中良，管金伟，刘强，李大鹏，郑文峰，胡峰.基于单积分球技术的最佳样本厚度研究
及油茶籽油鉴伪[J].中国油脂,2025,50(2):.[GONG Zhongliang， GUAN Jinwei， LIU Qiang， LI Dapeng,
ZHENG Wenfeng, HU Feng.Optimal sample thickness based on single integrating sphere technique and authentication of oil-tea camellia seed oil[J].China Oils and Fats,2025,50(2):.] |
| 基于单积分球技术的最佳样本厚度研究
及油茶籽油鉴伪 |
| Optimal sample thickness based on single integrating sphere technique and authentication of oil-tea camellia seed oil |
| |
| DOI： |
| 中文关键词: 油茶籽油 鉴伪；单积分球技术；光学参数；蒙特卡罗算法；样本厚度；定性模型 |
| 英文关键词:oil-tea camellia seed oil authentication single integrating sphere technique optical parameters Monte Carlo algorithm sample thickness qualitative model |
| 基金项目:湖南省科技计划重点研发项目（2022NK2048）；湖南省教育厅科学项目（18B192，20A515）；湖南省自然科学基金（2020JJ4142）；湖南省林业杰青培养科研项目（XLK202108-7） |
|
| 摘要点击次数: 27 |
| 全文下载次数: 18 |
| 中文摘要: |
| 为减少单积分球技术采集光谱数据过程中的光损失，探索了油茶籽油检测时光损失量最小的样本厚度，并研究在此样本厚度下鉴别掺伪油茶籽油的能力。采用蒙特卡罗(MC)算法模拟单积分球技术下样本的测量，将模拟出的反射率(MR)和透射率(MT)作为实际值，单积分球采集的数据作为预测值，将实际值与预测值之间的平均相对误差(MRE)和均方根误差(RMSE)作为评价指标，确定最佳的样本厚度。按不同掺伪比例制备了230组样本，采集最佳样本厚度的掺伪油茶籽油光谱数据，结合逆向倍增(IAD)算法得到样本的吸收系数(μa)和约化散射系数(μs′)。将μa和μs′经过均值中心化预处理之后，利用 Kennard-Stone(K-S)算法以 7∶ 3 的比例将样本划分成训练集和测试集，分别建立基于支持向量机(SVM)和随机森林(RF)的多分类定性鉴别模型。结果表明：样本厚度为14 mm时，MR和MT的MRE和RMSE均相对较小；μa和μs′建立的SVM模型鉴别准确率分别为97.10%和95.65%，建立的RF模型鉴别准确率分别为98.55%和97.10%。因此，基于最佳样本厚度下的单积分球技术结合SVM和RF模型，可有效实现油茶籽油的快速鉴伪。 |
| 英文摘要: |
| To reduce light loss in the spectral data acquisition process using single integrating sphere technique,the sample thickness resulting in the minimum light loss for the detection of oil-tea camellia seed oil was investigated, and the ability of distinguishing adulterated oil-tea camellia seed oil at this optimal sample thickness was researched. The Monte Carlo (MC) algorithm was used to simulate the measurement of samples under the single integrating sphere technique, and the simulated reflectance (MR) and transmittance (MT) were taken as the actual values,the data collected by the single integrating sphere were taken as the predicted values, and the mean relative error (MRE) and the root mean square error (RMSE) between the actual values and the predicted values were taken as the evaluation indexes to determine the optimal sample thickness. Two hundred and thirty sets of samples were prepared with different adulteration ratios, and the spectral data of adulterated oil-tea camellia seed oil with the optimal sample thickness were collected to obtain the absorption coefficients (μa) and approximate scattering coefficients (μs′) of the samples by combining with the inverse doubling (IAD) algorithm. After pre-processing the μa and μs′ by mean centering, the samples were divided into training and test sets using the Kennard-Stone (K-S) algorithm in the ratio of 7∶ 3. Multiclassification qualitative identification models based on support vector machine (SVM) and random forest (RF) were established, respectively. The results showed that the MRE and RMSE of both MR and MT were relatively small when the sample thickness was 14 mm.The discriminative accuracies of the SVM models established for μa and μs′ were 97.10% and 95.65%, respectively, and the discriminative accuracies of the RF models established were 98.55% and 97.10%, respectively. Therefore, based on the single integrating sphere technique under the optimal sample thickness, combined with SVM and RF models, the fast authentication of oil-tea camellia seed oil can be effectively realized. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
